| { | |
| "best_metric": 9.899702823179792, | |
| "best_model_checkpoint": "./data/finetuningRuns/whisper-lg-el-intlv-xs/checkpoint-4000", | |
| "epoch": 9.950248756218905, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.61e-07, | |
| "loss": 0.5163, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 3.36e-07, | |
| "loss": 0.4817, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 5.11e-07, | |
| "loss": 0.3656, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 6.86e-07, | |
| "loss": 0.2421, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 8.61e-07, | |
| "loss": 0.2035, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.036e-06, | |
| "loss": 0.1765, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.211e-06, | |
| "loss": 0.1664, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.379e-06, | |
| "loss": 0.1451, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.554e-06, | |
| "loss": 0.1496, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.7289999999999999e-06, | |
| "loss": 0.1385, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.9040000000000001e-06, | |
| "loss": 0.1285, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 2.079e-06, | |
| "loss": 0.1227, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.254e-06, | |
| "loss": 0.1184, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.429e-06, | |
| "loss": 0.113, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.604e-06, | |
| "loss": 0.1087, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.779e-06, | |
| "loss": 0.1128, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.954e-06, | |
| "loss": 0.0815, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 3.129e-06, | |
| "loss": 0.0883, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.3039999999999996e-06, | |
| "loss": 0.0762, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 3.479e-06, | |
| "loss": 0.0804, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 3.4918947368421053e-06, | |
| "loss": 0.0724, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 3.482684210526316e-06, | |
| "loss": 0.0726, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 3.4734736842105263e-06, | |
| "loss": 0.0759, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 3.464263157894737e-06, | |
| "loss": 0.0809, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 3.4550526315789474e-06, | |
| "loss": 0.0745, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 3.4458421052631577e-06, | |
| "loss": 0.0671, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 3.4366315789473685e-06, | |
| "loss": 0.0679, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 3.427421052631579e-06, | |
| "loss": 0.0728, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 3.4182105263157896e-06, | |
| "loss": 0.061, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 3.409e-06, | |
| "loss": 0.0619, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 3.39978947368421e-06, | |
| "loss": 0.0672, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 3.3909473684210528e-06, | |
| "loss": 0.0611, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 3.381736842105263e-06, | |
| "loss": 0.0409, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 3.372526315789474e-06, | |
| "loss": 0.032, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 3.363315789473684e-06, | |
| "loss": 0.0308, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 3.354105263157895e-06, | |
| "loss": 0.0318, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 3.3448947368421052e-06, | |
| "loss": 0.035, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 3.3356842105263156e-06, | |
| "loss": 0.0353, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 3.3264736842105263e-06, | |
| "loss": 0.0301, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 3.3172631578947366e-06, | |
| "loss": 0.0311, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "eval_loss": 0.1808706372976303, | |
| "eval_runtime": 1982.2643, | |
| "eval_samples_per_second": 0.856, | |
| "eval_steps_per_second": 0.107, | |
| "eval_wer": 10.549777117384844, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 3.3080526315789474e-06, | |
| "loss": 0.0309, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 3.2988421052631577e-06, | |
| "loss": 0.0364, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 3.2896315789473685e-06, | |
| "loss": 0.0323, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 3.2804210526315788e-06, | |
| "loss": 0.0322, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 3.271210526315789e-06, | |
| "loss": 0.0304, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 3.2620000000000003e-06, | |
| "loss": 0.0299, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.2527894736842106e-06, | |
| "loss": 0.0272, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.243578947368421e-06, | |
| "loss": 0.0241, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 3.2343684210526317e-06, | |
| "loss": 0.0186, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 3.225157894736842e-06, | |
| "loss": 0.015, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 3.2159473684210527e-06, | |
| "loss": 0.0125, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 3.206736842105263e-06, | |
| "loss": 0.0127, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 3.197526315789474e-06, | |
| "loss": 0.0144, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 3.188315789473684e-06, | |
| "loss": 0.0148, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 3.1791052631578945e-06, | |
| "loss": 0.0122, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 3.1698947368421052e-06, | |
| "loss": 0.0124, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 3.1606842105263155e-06, | |
| "loss": 0.0165, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 3.151842105263158e-06, | |
| "loss": 0.014, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.1426315789473684e-06, | |
| "loss": 0.0138, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 3.133421052631579e-06, | |
| "loss": 0.0142, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 3.1242105263157895e-06, | |
| "loss": 0.0136, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 3.115e-06, | |
| "loss": 0.0123, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 3.1057894736842106e-06, | |
| "loss": 0.0135, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 3.096578947368421e-06, | |
| "loss": 0.0135, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 3.0873684210526316e-06, | |
| "loss": 0.0102, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 3.078157894736842e-06, | |
| "loss": 0.0061, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 3.0689473684210523e-06, | |
| "loss": 0.0064, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 3.059736842105263e-06, | |
| "loss": 0.0068, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 3.0505263157894734e-06, | |
| "loss": 0.0056, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 3.041315789473684e-06, | |
| "loss": 0.0062, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 3.032105263157895e-06, | |
| "loss": 0.0071, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 3.022894736842105e-06, | |
| "loss": 0.0061, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 3.013684210526316e-06, | |
| "loss": 0.0069, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 3.0044736842105263e-06, | |
| "loss": 0.0067, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 2.995263157894737e-06, | |
| "loss": 0.0055, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 2.9860526315789473e-06, | |
| "loss": 0.0067, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 2.9768421052631577e-06, | |
| "loss": 0.0061, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 2.9676315789473684e-06, | |
| "loss": 0.0074, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 2.9584210526315787e-06, | |
| "loss": 0.0087, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 2.9492105263157895e-06, | |
| "loss": 0.0074, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "eval_loss": 0.2470209300518036, | |
| "eval_runtime": 1972.083, | |
| "eval_samples_per_second": 0.86, | |
| "eval_steps_per_second": 0.108, | |
| "eval_wer": 10.280460624071322, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 2.94e-06, | |
| "loss": 0.0049, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 2.9307894736842106e-06, | |
| "loss": 0.0035, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 2.921578947368421e-06, | |
| "loss": 0.0045, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 2.9123684210526316e-06, | |
| "loss": 0.0038, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 2.9031578947368424e-06, | |
| "loss": 0.0033, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 2.8939473684210527e-06, | |
| "loss": 0.0036, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 2.884736842105263e-06, | |
| "loss": 0.0039, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 2.8755263157894738e-06, | |
| "loss": 0.0035, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 2.866315789473684e-06, | |
| "loss": 0.0038, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 2.857105263157895e-06, | |
| "loss": 0.0039, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 2.847894736842105e-06, | |
| "loss": 0.0036, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 2.838684210526316e-06, | |
| "loss": 0.0043, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 2.8294736842105262e-06, | |
| "loss": 0.0049, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 2.8202631578947366e-06, | |
| "loss": 0.0041, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 2.8110526315789473e-06, | |
| "loss": 0.004, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 2.8018421052631576e-06, | |
| "loss": 0.0036, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 2.7926315789473684e-06, | |
| "loss": 0.003, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 2.783421052631579e-06, | |
| "loss": 0.0026, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 2.7742105263157895e-06, | |
| "loss": 0.0019, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 2.765e-06, | |
| "loss": 0.0023, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 2.7557894736842105e-06, | |
| "loss": 0.0029, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 2.7465789473684213e-06, | |
| "loss": 0.0032, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 2.7373684210526316e-06, | |
| "loss": 0.0028, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 2.728157894736842e-06, | |
| "loss": 0.0025, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 2.7189473684210527e-06, | |
| "loss": 0.0028, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 2.709736842105263e-06, | |
| "loss": 0.0025, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 2.7005263157894738e-06, | |
| "loss": 0.0024, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 2.691315789473684e-06, | |
| "loss": 0.0032, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 2.6821052631578944e-06, | |
| "loss": 0.0031, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 2.672894736842105e-06, | |
| "loss": 0.0032, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 2.6636842105263155e-06, | |
| "loss": 0.0034, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 2.6544736842105266e-06, | |
| "loss": 0.0026, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 2.645263157894737e-06, | |
| "loss": 0.003, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 2.6360526315789473e-06, | |
| "loss": 0.003, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 2.626842105263158e-06, | |
| "loss": 0.0023, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 2.6176315789473684e-06, | |
| "loss": 0.0024, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 2.608421052631579e-06, | |
| "loss": 0.0019, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 2.5992105263157894e-06, | |
| "loss": 0.0021, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 2.5899999999999998e-06, | |
| "loss": 0.0019, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 2.5807894736842105e-06, | |
| "loss": 0.0019, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "eval_loss": 0.30083394050598145, | |
| "eval_runtime": 1980.8887, | |
| "eval_samples_per_second": 0.856, | |
| "eval_steps_per_second": 0.107, | |
| "eval_wer": 10.029717682020802, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 2.571578947368421e-06, | |
| "loss": 0.0022, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 2.5623684210526316e-06, | |
| "loss": 0.0022, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 2.553157894736842e-06, | |
| "loss": 0.0021, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 2.5439473684210522e-06, | |
| "loss": 0.0027, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "learning_rate": 2.5347368421052634e-06, | |
| "loss": 0.0023, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 2.5255263157894737e-06, | |
| "loss": 0.0026, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 7.9, | |
| "learning_rate": 2.5163157894736845e-06, | |
| "loss": 0.0022, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 2.507105263157895e-06, | |
| "loss": 0.0023, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 2.497894736842105e-06, | |
| "loss": 0.0023, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 2.488684210526316e-06, | |
| "loss": 0.0018, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 2.479473684210526e-06, | |
| "loss": 0.0015, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 2.470263157894737e-06, | |
| "loss": 0.0016, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 2.4610526315789473e-06, | |
| "loss": 0.0016, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 2.4518421052631576e-06, | |
| "loss": 0.0016, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 2.4426315789473684e-06, | |
| "loss": 0.0014, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 2.4334210526315787e-06, | |
| "loss": 0.0015, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 2.4242105263157894e-06, | |
| "loss": 0.0016, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 2.4149999999999997e-06, | |
| "loss": 0.0019, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 2.4057894736842105e-06, | |
| "loss": 0.0015, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 2.3965789473684212e-06, | |
| "loss": 0.0022, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 8.77, | |
| "learning_rate": 2.3873684210526316e-06, | |
| "loss": 0.0018, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 2.3781578947368423e-06, | |
| "loss": 0.0027, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 2.3689473684210526e-06, | |
| "loss": 0.0016, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 2.359736842105263e-06, | |
| "loss": 0.0016, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 2.3505263157894737e-06, | |
| "loss": 0.0015, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 2.341315789473684e-06, | |
| "loss": 0.001, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 9.14, | |
| "learning_rate": 2.3321052631578948e-06, | |
| "loss": 0.0011, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 9.2, | |
| "learning_rate": 2.322894736842105e-06, | |
| "loss": 0.0011, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 2.3136842105263154e-06, | |
| "loss": 0.0012, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 2.304473684210526e-06, | |
| "loss": 0.0012, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 2.2952631578947365e-06, | |
| "loss": 0.0012, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 2.2860526315789477e-06, | |
| "loss": 0.0011, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 2.276842105263158e-06, | |
| "loss": 0.0009, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 2.2676315789473683e-06, | |
| "loss": 0.0009, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 9.64, | |
| "learning_rate": 2.258421052631579e-06, | |
| "loss": 0.0014, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 9.7, | |
| "learning_rate": 2.2492105263157894e-06, | |
| "loss": 0.0008, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 2.24e-06, | |
| "loss": 0.0015, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 9.83, | |
| "learning_rate": 2.2307894736842105e-06, | |
| "loss": 0.0009, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 9.89, | |
| "learning_rate": 2.221578947368421e-06, | |
| "loss": 0.0012, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 2.2123684210526316e-06, | |
| "loss": 0.0011, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "eval_loss": 0.29131078720092773, | |
| "eval_runtime": 1980.7982, | |
| "eval_samples_per_second": 0.856, | |
| "eval_steps_per_second": 0.107, | |
| "eval_wer": 9.899702823179792, | |
| "step": 4000 | |
| } | |
| ], | |
| "max_steps": 10000, | |
| "num_train_epochs": 25, | |
| "total_flos": 2.716512394272768e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |